用可复用的推理记忆，让大模型在压缩思考过程时仍保持高准确率。
Memory Augmentation Unlocks Efficient Chain-of-Thought Reasoning
- 从历史推理中提炼通用模式，构建可重用的推理记忆
- 在GSM8K等任务上提升21.4~29.5分，推理速度提升1.14~1.49倍
- 无需训练，兼容多种压缩方式，适合高效推理场景
大语言模型常依赖链式思维（CoT）解决复杂任务，但冗长的推理过程带来显著推理开销。链式思维压缩虽能缩短生成时间，但激进压缩可能破坏逻辑连贯性并降低性能。本文提出‘上下文-生成替代定律’，即显式推理上下文可替代部分解码时的生成内容。基于此，我们设计无训练的内存增强压缩框架，从历史推理轨迹中构建可复用的推理记忆，并在预填充阶段作为支架调用。这些记忆不直接使用原始示例，而是总结通用推理模式、关键约束与核心操作，以补偿压缩带来的信息损失。实验表明，该方法在数学推理、复杂推理及科学问答任务上持续提升提示引导的链式草稿（CoD）压缩性能，在GSM8K、MATH、BBH和MMLU-Sci上分别获得21.4、28.0、29.5和6.61分的准确率提升，同时实现1.14~1.49倍的延迟加速。该方法还兼容词级、推理轨迹级及推理状态压缩机制。
原文摘要 · Abstract (English)
Large language models often rely on Chain-of-Thought (CoT) reasoning to solve complex tasks, but verbose reasoning traces introduce substantial inference overhead. CoT compression shortens generation, yet aggressive compression may disrupt logical coherence and degrade performance. We formalize this trade-off as the Context-Generation Substitution Law, where explicit reasoning context substitutes for part of decode-time generation. Based on this principle, we propose Memory-Augmented Compression, a training-free framework that constructs reusable reasoning memories from historical traces and retrieves them as prefill-side scaffolds. Rather than using raw demonstrations, these memories summarize reusable reasoning patterns, key constraints, and critical operations to compensate for information lost during compression. Experiments show that Memory consistently improves prompt-based Chain-of-Draft (CoD) compression across mathematical reasoning, complex reasoning, and science question answering tasks, yielding accuracy gains of 21.4, 28.0, 29.5, and 6.61 points over CoD on GSM8K, MATH, BBH, and MMLU-Sci, while achieving a 1.14-1.49x latency speedup latency speedup over standard CoT. Memory is also compatible with token-level, reasoning-trace-level, and inference-state compression mechanisms.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。